Stores that sell to the United States picked up a new path before checkout this week. On October 15, Google opened Universal Cart in Google Search and the Gemini app, in the U.S. only. A shopper can save a Steve Madden bag from AI Mode in Search, then a Quince sweater and Ulta Beauty lip gloss from the Gemini app, and those items land in the same cart.
Keep one sentence: this changes the path before anyone pays. It is not a published conversion-rate result. Google did not give a before-and-after figure for clicks or orders. The merchant still sends the confirmation and the support, and the merchant stays the seller of record.
Four checks decide what you do this week: whether the shopper is in the U.S., whether they are still inside Google or already in your checkout, whose name is on the order, and whether any public note actually contains a conversion number. The notes are the October 15 launch post, the May 19 I/O preview, and the UCP Cart API guide. None of the percentages below come from a store’s admin.
After Universal Cart launched, how should you read traffic and conversion this week?
| What you might see this week | Can you call it a conversion-rate change? | What to do |
|---|---|---|
| A U.S. shopper saves items from several stores in one cart, in Search or Gemini | No | They are still inside Google. That is discovery, not an order |
| The cart flags a price drop, a restock, a promo code, or a card perk | No | An alert is not a sale. Do not move ad budget because of it |
| They pay with Google Pay in the cart, or land on your site with the cart already filled | The path changed | Product-page sessions and orders can diverge. Match orders in the merchant admin, not the landing-page rate alone |
| Your shoppers are mostly outside the United States | No | The October 15 scope does not cover them. Do not attribute this week’s swing to Universal Cart |
Where does Universal Cart work right now?
The May 19 I/O preview talked about the cart across Search, Gemini, YouTube, and Gmail. It said the U.S. Search and Gemini apps would open in the summer, with YouTube and Gmail later. The October 15 launch post names only two places: Google Search in the United States, and the Gemini app. The Verge’s same-day report says Gmail and YouTube still have no timeline.
| When | What the public note says | Treat it as live? |
|---|---|---|
| May 19, 2026 | I/O preview. The cart runs on Gemini models and can watch price, stock, promos, and payment perks. The examples include a parts-compatibility warning | Still a preview |
| October 15, 2026 | Google Search in the U.S., including AI Mode, and the Gemini app. One cart across retailers, price and restock alerts, promo codes and card perks, Google Pay or a handoff to the merchant cart | Those four are in the launch post |
| YouTube and Gmail | May said they would follow. The October post does not list them as open | Do not make a channel for them yet |
| Outside the United States | Both Google posts limit this rollout to the U.S. | Not covered |
After someone saves an item, do they still open your product page first?
Not always. The launch post describes two ways to finish. One is Google Pay inside the cart. The stores named as places to try it include Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and many Shopify merchants. The other is a transfer into the merchant’s own cart, so the shopper pays on your site. Either way, the merchant sends the confirmation and the support, and the merchant stays the seller of record.
That is why “will AI search steal the product page” needs two answers. If the shopper pays inside Google, product-page sessions can fall while the order still shows in the merchant admin. If they are transferred back, you do get a session, but it often starts on a cart that is already filled, not on the product detail. Neither case is “search traffic arrives as before and converts on the product page.”
| Checkout path | Product page first? | Where the order sits |
|---|---|---|
| Saved in Google, not paid yet | No | There is no order yet |
| Paid with Google Pay in the cart | Can skip it | The merchant is still the seller. Match the order in the merchant admin |
| Transferred to the merchant cart, then paid | They enter the site, often on the cart | Same merchant order. The landing URL is not the product URL |
Which number in the traffic report moves first?
The cell that moves first is usually not “sitewide conversion rate.” It is the page a session starts on. Someone who only saves inside Google never creates a product-page session. Someone who is sent back to your site creates a cart landing. A price-drop or restock alert can bring them back days later, and that return visit should not be folded into the conversion rate of the day they first saved the item.
Google has not published how far each of those cells moved. You can only read your own admin, and you have to keep U.S. traffic apart from every other country. Mixed together, an ordinary swing outside the U.S. will exaggerate or cancel this launch.
- Run a product query as a U.S. shopper and see whether your item is still in Google shopping results. If it is absent, you are not on this path yet.
- Keep three columns for the U.S.: product-page sessions, cart-landing sessions, and orders in the merchant admin. Do not collapse them into one rate.
- If cart transfer is already live, tag that landing URL and read the completion rate from landing to payment on its own.
- Log price alerts and promo-code displays as alerts, not as orders.
Did the conversion rate rise, or did checkout move?
The public notes cannot answer “by how much it rose.” They can answer where checkout is allowed to happen: inside Google’s cart, or inside your cart after the items were placed there in advance. A site conversion rate uses sessions that reached the site as the denominator. When the denominator changes, the rate changes, even if the merchant’s order count does not.
| Easy to treat as fact | Do the public notes support it? | What to do in the report |
|---|---|---|
| AI search has already raised ecommerce conversion | No | There is no lift figure. Do not reset the year’s conversion target |
| Google became the seller | No | The merchant stays the seller of record. Match orders and support on the merchant side |
| A lower product-page conversion rate means worse traffic | Not necessarily | Check whether the order moved to Google Pay or to a prefilled cart |
| The cart stays in sync both ways with your site cart | No | This release only creates a cart. It does not update, delete, or sync. Do not build inventory write-back for it |
Do you need the UCP Cart API this week?
Only if you want to receive a prefilled cart handed back from Google. The guide says the Cart API exists starting with UCP version 2026-04-08. The capability name is dev.ucp.shopping.cart. This release is CreateCart only: your system accepts the line items and returns a continue_url, and the shopper lands on a cart that already holds those items. It does not update, delete, or sync with Google, and it does not take payment.
Paying with Google Pay inside the cart is the other checkout experience in the launch post. It is not this create-cart call. Nike, Sephora, and Target are examples Google named, not a complete list of stores that appear automatically once an API is connected. The guide also does not say the conversion rate rises by any number of points after you connect it.
| Capability you want | Does this Cart API provide it? | What to do this week |
|---|---|---|
| Send a U.S. shopper from Search or Gemini to a prefilled cart | Yes, one-way create | Implement CreateCart from the guide, and record continue_url as its own landing |
| Update, delete, or sync with the on-site cart | No | Do not build two-way inventory for this release |
| Take Google Pay inside Google’s cart | Not this API | That is checkout. Judge it separately. The merchant list is whatever Google named at the time |
| Use a published lift number to justify the sprint | That number does not exist | The reason to schedule it is “U.S. shoppers can arrive this way,” not “conversion has already risen” |
If the store does not sell to the U.S., do you change attribution?
No. The October 15 scope is Search and the Gemini app in the United States. When shoppers are in other countries, this week’s session swings and product-page rate swings still have no new path to explain them. Writing an ordinary swing down as “AI search already changed conversion” means next month’s table will not match.
The only preparation worth doing is to make sure the order’s country field can separate U.S. orders, in case those products later appear in U.S. shopping results. If the field cannot be split, you still will not be able to tell a path change from a demand change when the product reaches your market.
Questions people search
When did Universal Cart launch, and which countries have it?
Google’s October 15, 2026 post says it opened that day in Google Search and the Gemini app in the United States. The May 19 preview had said those two surfaces would open in the summer. YouTube and Gmail are not on the October list of open surfaces.
Did Google publish a conversion or traffic lift?
No. The launch post describes what the cart can do. It does not compare clicks or orders before and after. “More than a billion” shopping moments a day is the May description of shopping on Google, not a score for this feature.
If someone pays with Google Pay in the cart, is the brand still the seller?
Yes. Google says the merchant sends the confirmation and the support, and the merchant remains the seller of record. Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and some Shopify merchants are examples, not a full list.
Is the Cart API the same thing as checkout?
No. This Cart API release only creates a cart in one direction and returns a continue_url. It does not take payment, and it does not update, delete, or sync. Google Pay checkout is a separate experience.
If product-page traffic falls, did conversion get worse?
The product page alone cannot answer that. Split U.S. product-page sessions, prefilled-cart landings, and orders in the merchant admin. If the orders are still there and the product-page sessions are not, read it as checkout moving first.
What each kind of store should do this week
If your shoppers are in the United States and the products already appear in Google shopping results, keep product-page sessions, prefilled-cart landings, and merchant-admin orders in separate columns. Do not reset the year’s conversion target against a lift that has not been published.
If people can pay with Google Pay inside the cart, match orders in the merchant admin. When the product-page conversion rate falls on its own, check whether that order skipped the product page.
If people come back to your site to pay, schedule CreateCart only for that prefilled cart, and treat continue_url as its own landing. Do not build two-way sync for this release.
If the store does not sell to the U.S., or the products are not in shopping results yet, leave the attribution rules where they are. When the scope includes your market and the products actually appear, use the same three columns.
Related reading
- What changed in Gemini Agent and tool calling in 2026: Universal Cart also shows up in the Gemini app. That piece is about the developer interface. Use it to keep an action inside a chat separate from a store order, instead of folding both into one conversion rate.
- How to read the 2026 ChatGPT feature changes: another product’s update note is also not a sales lift. Read them side by side, and do not reset the year’s target on launch day.
